Indoor robot localization combining feature clustering with wireless sensor network

Abstract

Indoor robot localization is an indispensable ingredient for robots to perform autonomous services because GPS (Global Position System) information is not available. Natural features are usually used to implement this task, but it is difficult to solve the problem of localization robustness. A solution is proposed combining feature clustering and wireless sensor network to improve the effectiveness of robot localization: firstly, the SIFT (scalable invariable feature transform) features are extracted with feature clustering algorithm to estimate the robot position; secondly, the wireless sensor network is constructed to localize the robot from another independent way; finally, EKF (extended Kalman filter) is utilized to fuse the two kinds of localization results. The experiments demonstrate that this proposed method is effective and robust.